Experiment 224
A6.1 (Stage 2, step 4, issue #84), the candidate 0222 left (0223 is a mistaken duplicate of 0222: the new code was not written before the run). 0222 judged K4's saved readings of 0034, 0035 and 0048 against 0216's regenerated copies: in the cell 0034 at 27-29 K4's best (8 of 100) was above every one of 10 copies (4-7), the only one of 4 cells where it led (chance 0.36). Pre-registered rule (program.md): a candidate is replicated on that cell alone against fresh copies, enough for the corrected p (0.05 / 4: 80 copies all below K4), before anything else. Fresh plain shuffles (seed 4222), 0034's frozen sweep unchanged at periods 27, 28, 29 (the cell's three readings per text: 24 conventions, 20 restarts each), eight at a time, judged after each batch.
What the method does with the 24 known letters.
One old search, rerun on scrambled copiesa shuffled control The same experiment run on the carved letters shuffled into a random order. of the carving, gave readings the judgethe judge A language model that scores how much of a reading is English, from 0 to 100, allowing misspellings and X between sentences. scored a little below the carving's reading (7 against 8 out of 100, both near the bottom of the scale). That is probably luck. The honest check is to scramble again and again, run the same search, and see how soon a scrambled copy does as well.
Words in the margin are defined on the glossary page.
The carved letters go in at the top; the reading comes out at the bottom.
In 0222 K4's readings at key lengths 27-29 scored 8 out of 100 with the judge, every copy 7 or less. Luck is tested by more copies.
Fresh scrambles through 0034's frozen search at 27-29, judged in batches of eight; 80 in a row below 8 would be a flag.
Copy 6 of the first batch scored 8: not a flag; A6.1's cell closes. Recorded as kept by mistake: gate 4 was given only the 8 fresh copies, and a K4 reading that fails against its own search's copies passed (issue #94); 0225 reruns it.
What this experiment was built on, from the lineage page: 15 experiments before it, back to the root of its line. Its number is in bold.
Laid out as the copper carries K4: OBKR ends the line that finishes K3, then three lines of 31. Left, as carved, the 24 known places in accent. Right, the candidate: a known place it gets right is an accent cell, one it misses is outlined. This is the lowest-loss reading that passed the original gates.
in: the copper, always the same
OBKR ends the line that finishes K3, then three lines of 31
out: the candidate
checked: the known letters in place, and the reading turns back into the copper
The loss of the best candidate, part by part. Zero is solved.
| Known letters missed | 0 of 24, ten points each | 0.00 |
| English | the 73 free letters score -6.71; English text scores -3.6 | 3.11 |
| Key | 151 bits to write it down (key words are cheap, arbitrary letters 4.7 bits each); each 60 beyond 60 costs a point | 1.51 |
| Loss | of 240, the worst possible | 4.62 1.9% |
Redo the method on a few letters with pencil and paper.
The same steps run by the code on the real letters.
0034's reading of K4 at key length 27 (3 free key slots climbed):
reading ZWAWFAFIWGMGOGMDROTMAEASTNORTHEASTISTCDAPPUDSFROKLPDGBCUMLDWLBYBERLINCLOCKADHARIUHVWCQFISIQMNGDJK
judge sees ZWAWFAFIWGMGOGMDROTMA#############ISTCDAPPUDSFROKLPDGBCUMLDWLBY###########ADHARIUHVWCQFISIQMNGDJK
K4's best of its three readings (27, 28, 29) scored 8; a fresh copy's three readings by the same search must reach 8.
Family, key space, the constraint under test, and the clues it reads.
A6 (knowledge/12-assumptions.md). Cipher: one periodic key on the standard or KRYPTOS tables, 24 conventions, key lengths 27-29 (C1(c), C27(c): the K1/K2 tableau system at the only lengths 1-52 the known letters leave alive for it; 0005), the 3-5 free key slots completed by a quadgram climb (0034). Statistic: the judge's best score (j3, Sonnet 5.5) over a text's three readings. K4: 8 ('? LEVES SIT MILAN BAC ? AT ? OF ?'). With 3-5 free slots each reading is fixed but for 9-15 letters, so a score of 8 means a few short words, inside what the bench calls chance (20 = a few short words). Sequential: a fresh copy at or above 8 ends it (not a flag); 80 copies all below K4 is a flag (p < 0.0125 corrected for the 4 cells of 0222).
Written before the run; the result decides it.
A fresh copy scores 8 or more within the first two batches (16 copies): the 10 copies of 0222 already reached 7 twice, and the judge's scores at this end of the scale move in steps of 2-4. Not a flag. Zero candidates pass gate 4.
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
All 3 candidates, lowest loss first.
saved reading of 0034: period 29 key HYHMKBNENTXBQRMAEROOOXYHANZEF (5 free slots climbed); kry/kry/std/beau
saved reading of 0034: period 28 key NZVDMORFNOEEADIAGKSSSUELRXLO (4 free slots climbed); kry/std/std/vig
saved reading of 0034: period 27 key EFHYHMKUKBNENTXBQRMAEXYHANZ (3 free slots climbed); kry/kry/std/beau
not a flag: 8 fresh copies, first at or above K4's 8: copy 6; copies' best scores [3, 4, 4, 4, 4, 4, 8, 4]
What was searched, as the harness recorded it.
OGNJPNCPUSGEDHWJPKFBIEASTNORTHEASTCERSUCWUNHETISNORELPFKEORDSFIBERLINCLOCKERMEGTDGCEFQVYONCLTKOWIXMLSSEWRYAUAISHQZGRXEEASTNORTHEASTONCYCOYESDLWNEKJHOITMVCRHYYERBERLINCLOCKOFYSHUDOLLHRWGOPJFOGRIBTBYFBDVYMDXYPPCVEIOBOEASTNORTHEASTTGUDEIXCYKWKINGHEITYRCABNUXBNBERLINCLOCKETEDELYGXTKPGBIGOVVAZZHMHFIKPXKAWGLTOMVEAHAVEASTNORTHEASTEEYAPENMBUYHAGECUPQCHHHKRASGMBERLINCLOCKFORDTELVXTCOMVQBFOEYEXZISYLCZGCRVSUSOJNBRNGOEASTNORTHEASTMIXADXRSHZSTBLAGLAGHEGGGGJRAPBERLINCLOCKEEUUMKPBVQFXRQALLHIMQLXTLSJNAILLWEZELMAENXWREASTNORTHEASTITIIOIHLXJXRHEJBYPLAYIHIFUZAIBERLINCLOCKSVRMARZKTENLEFUMBOPEMYXBMHIHAVEBELDPUSBMPCJLEASTNORTHEASTAKIKNSEVNUPJBTVKIMIWJEEIQFENFBERLINCLOCKREFOKYMZFKNRDHNTYPEISWXTYCZIQUASAGFGXHOXZYSTEASTNORTHEASTRONOMYABAEFJEAHUAIIPUETCIPLDWBERLINCLOCKHOFMWJUKTNKAKPRVOAYYZJMHow this experiment sits among the others is on the lineage page; its family's open hypotheses are on the families page.